Probabilistic post-processing of short to medium range temperature forecasts: Implications for heatwave prediction in...

Area of Science:

• Meteorology and Climate Science
• Atmospheric Science
• Environmental Science

Background:

• Accurate air temperature forecasts are crucial for managing thermal disasters like heat strokes.
• Numerical Weather Prediction (NWP) models often have biases requiring post-processing.
• Research on probabilistic post-processing techniques (PPTs) for temperature forecasts in India is limited.

Purpose of the Study:

• To evaluate Nonhomogeneous Gaussian Regression (NGR) and Bayesian Model Averaging (BMA) for improving NWP temperature forecasts in India.
• To assess the impact of PPTs on heatwave prediction skills across India.
• To identify the most effective PPT for Indian temperature forecasting.

Main Methods:

• Utilized daily temperature forecasts from ECMWF and GEFS NWP models.
• Applied probabilistic post-processing techniques: Nonhomogeneous Gaussian Regression (NGR) and Bayesian Model Averaging (BMA).
• Evaluated forecast performance across the Indian subcontinent, including Himalayan regions, and assessed heatwave prediction skills.

Main Results:

• Probabilistic PPTs significantly outperformed traditional methods for temperature forecasting across India at all lead times.
• NGR demonstrated superior performance compared to BMA and other PPTs in the Indian region.
• While probabilistic techniques improved forecasts in low-skill regions like the Himalayas, they did not achieve skillful forecasts.
• Post-processing maximum temperature (Tmax) with NGR substantially enhanced heatwave prediction accuracy in vulnerable areas.

Conclusions:

• Probabilistic post-processing, particularly NGR, offers a significant improvement for temperature forecasting and heatwave prediction in India.
• The findings support the development of enhanced heatwave early warning systems for India.
• Further research may be needed to improve forecast skill in regions with inherently low raw forecast accuracy, such as the Himalayas.